1/* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * RAW - implementation of IP "raw" sockets. 7 * 8 * Version: $Id: raw.c,v 1.1.1.1 2007/08/03 18:53:51 Exp $ 9 * 10 * Authors: Ross Biro 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 12 * 13 * Fixes: 14 * Alan Cox : verify_area() fixed up 15 * Alan Cox : ICMP error handling 16 * Alan Cox : EMSGSIZE if you send too big a packet 17 * Alan Cox : Now uses generic datagrams and shared 18 * skbuff library. No more peek crashes, 19 * no more backlogs 20 * Alan Cox : Checks sk->broadcast. 21 * Alan Cox : Uses skb_free_datagram/skb_copy_datagram 22 * Alan Cox : Raw passes ip options too 23 * Alan Cox : Setsocketopt added 24 * Alan Cox : Fixed error return for broadcasts 25 * Alan Cox : Removed wake_up calls 26 * Alan Cox : Use ttl/tos 27 * Alan Cox : Cleaned up old debugging 28 * Alan Cox : Use new kernel side addresses 29 * Arnt Gulbrandsen : Fixed MSG_DONTROUTE in raw sockets. 30 * Alan Cox : BSD style RAW socket demultiplexing. 31 * Alan Cox : Beginnings of mrouted support. 32 * Alan Cox : Added IP_HDRINCL option. 33 * Alan Cox : Skip broadcast check if BSDism set. 34 * David S. Miller : New socket lookup architecture. 35 * 36 * This program is free software; you can redistribute it and/or 37 * modify it under the terms of the GNU General Public License 38 * as published by the Free Software Foundation; either version 39 * 2 of the License, or (at your option) any later version. 40 */ 41 42#include <linux/types.h> 43#include <asm/atomic.h> 44#include <asm/byteorder.h> 45#include <asm/current.h> 46#include <asm/uaccess.h> 47#include <asm/ioctls.h> 48#include <linux/stddef.h> 49#include <linux/slab.h> 50#include <linux/errno.h> 51#include <linux/aio.h> 52#include <linux/kernel.h> 53#include <linux/spinlock.h> 54#include <linux/sockios.h> 55#include <linux/socket.h> 56#include <linux/in.h> 57#include <linux/mroute.h> 58#include <linux/netdevice.h> 59#include <linux/in_route.h> 60#include <linux/route.h> 61#include <linux/skbuff.h> 62#include <net/dst.h> 63#include <net/sock.h> 64#include <linux/gfp.h> 65#include <linux/ip.h> 66#include <linux/net.h> 67#include <net/ip.h> 68#include <net/icmp.h> 69#include <net/udp.h> 70#include <net/raw.h> 71#include <net/snmp.h> 72#include <net/tcp_states.h> 73#include <net/inet_common.h> 74#include <net/checksum.h> 75#include <net/xfrm.h> 76#include <linux/rtnetlink.h> 77#include <linux/proc_fs.h> 78#include <linux/seq_file.h> 79#include <linux/netfilter.h> 80#include <linux/netfilter_ipv4.h> 81 82struct hlist_head raw_v4_htable[RAWV4_HTABLE_SIZE]; 83DEFINE_RWLOCK(raw_v4_lock); 84 85static void raw_v4_hash(struct sock *sk) 86{ 87 struct hlist_head *head = &raw_v4_htable[inet_sk(sk)->num & 88 (RAWV4_HTABLE_SIZE - 1)]; 89 90 write_lock_bh(&raw_v4_lock); 91 sk_add_node(sk, head); 92 sock_prot_inc_use(sk->sk_prot); 93 write_unlock_bh(&raw_v4_lock); 94} 95 96static void raw_v4_unhash(struct sock *sk) 97{ 98 write_lock_bh(&raw_v4_lock); 99 if (sk_del_node_init(sk)) 100 sock_prot_dec_use(sk->sk_prot); 101 write_unlock_bh(&raw_v4_lock); 102} 103 104struct sock *__raw_v4_lookup(struct sock *sk, unsigned short num, 105 __be32 raddr, __be32 laddr, 106 int dif) 107{ 108 struct hlist_node *node; 109 110 sk_for_each_from(sk, node) { 111 struct inet_sock *inet = inet_sk(sk); 112 113 if (inet->num == num && 114 !(inet->daddr && inet->daddr != raddr) && 115 !(inet->rcv_saddr && inet->rcv_saddr != laddr) && 116 !(sk->sk_bound_dev_if && sk->sk_bound_dev_if != dif)) 117 goto found; /* gotcha */ 118 } 119 sk = NULL; 120found: 121 return sk; 122} 123 124/* 125 * 0 - deliver 126 * 1 - block 127 */ 128static __inline__ int icmp_filter(struct sock *sk, struct sk_buff *skb) 129{ 130 int type; 131 132 if (!pskb_may_pull(skb, sizeof(struct icmphdr))) 133 return 1; 134 135 type = icmp_hdr(skb)->type; 136 if (type < 32) { 137 __u32 data = raw_sk(sk)->filter.data; 138 139 return ((1 << type) & data) != 0; 140 } 141 142 /* Do not block unknown ICMP types */ 143 return 0; 144} 145 146/* IP input processing comes here for RAW socket delivery. 147 * Caller owns SKB, so we must make clones. 148 * 149 * RFC 1122: SHOULD pass TOS value up to the transport layer. 150 * -> It does. And not only TOS, but all IP header. 151 */ 152int raw_v4_input(struct sk_buff *skb, struct iphdr *iph, int hash) 153{ 154 struct sock *sk; 155 struct hlist_head *head; 156 int delivered = 0; 157 158 read_lock(&raw_v4_lock); 159 head = &raw_v4_htable[hash]; 160 if (hlist_empty(head)) 161 goto out; 162 sk = __raw_v4_lookup(__sk_head(head), iph->protocol, 163 iph->saddr, iph->daddr, 164 skb->dev->ifindex); 165 166 while (sk) { 167 delivered = 1; 168 if (iph->protocol != IPPROTO_ICMP || !icmp_filter(sk, skb)) { 169 struct sk_buff *clone = skb_clone(skb, GFP_ATOMIC); 170 171 /* Not releasing hash table! */ 172 if (clone) 173 raw_rcv(sk, clone); 174 } 175 sk = __raw_v4_lookup(sk_next(sk), iph->protocol, 176 iph->saddr, iph->daddr, 177 skb->dev->ifindex); 178 } 179out: 180 read_unlock(&raw_v4_lock); 181 return delivered; 182} 183 184void raw_err (struct sock *sk, struct sk_buff *skb, u32 info) 185{ 186 struct inet_sock *inet = inet_sk(sk); 187 const int type = icmp_hdr(skb)->type; 188 const int code = icmp_hdr(skb)->code; 189 int err = 0; 190 int harderr = 0; 191 192 /* Report error on raw socket, if: 193 1. User requested ip_recverr. 194 2. Socket is connected (otherwise the error indication 195 is useless without ip_recverr and error is hard. 196 */ 197 if (!inet->recverr && sk->sk_state != TCP_ESTABLISHED) 198 return; 199 200 switch (type) { 201 default: 202 case ICMP_TIME_EXCEEDED: 203 err = EHOSTUNREACH; 204 break; 205 case ICMP_SOURCE_QUENCH: 206 return; 207 case ICMP_PARAMETERPROB: 208 err = EPROTO; 209 harderr = 1; 210 break; 211 case ICMP_DEST_UNREACH: 212 err = EHOSTUNREACH; 213 if (code > NR_ICMP_UNREACH) 214 break; 215 err = icmp_err_convert[code].errno; 216 harderr = icmp_err_convert[code].fatal; 217 if (code == ICMP_FRAG_NEEDED) { 218 harderr = inet->pmtudisc != IP_PMTUDISC_DONT; 219 err = EMSGSIZE; 220 } 221 } 222 223 if (inet->recverr) { 224 struct iphdr *iph = (struct iphdr*)skb->data; 225 u8 *payload = skb->data + (iph->ihl << 2); 226 227 if (inet->hdrincl) 228 payload = skb->data; 229 ip_icmp_error(sk, skb, err, 0, info, payload); 230 } 231 232 if (inet->recverr || harderr) { 233 sk->sk_err = err; 234 sk->sk_error_report(sk); 235 } 236} 237 238static int raw_rcv_skb(struct sock * sk, struct sk_buff * skb) 239{ 240 /* Charge it to the socket. */ 241 242 if (sock_queue_rcv_skb(sk, skb) < 0) { 243 kfree_skb(skb); 244 return NET_RX_DROP; 245 } 246 247 return NET_RX_SUCCESS; 248} 249 250int raw_rcv(struct sock *sk, struct sk_buff *skb) 251{ 252 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb)) { 253 kfree_skb(skb); 254 return NET_RX_DROP; 255 } 256 nf_reset(skb); 257 258 skb_push(skb, skb->data - skb_network_header(skb)); 259 260 raw_rcv_skb(sk, skb); 261 return 0; 262} 263 264static int raw_send_hdrinc(struct sock *sk, void *from, size_t length, 265 struct rtable *rt, 266 unsigned int flags) 267{ 268 struct inet_sock *inet = inet_sk(sk); 269 int hh_len; 270 struct iphdr *iph; 271 struct sk_buff *skb; 272 int err; 273 274 if (length > rt->u.dst.dev->mtu) { 275 ip_local_error(sk, EMSGSIZE, rt->rt_dst, inet->dport, 276 rt->u.dst.dev->mtu); 277 return -EMSGSIZE; 278 } 279 if (flags&MSG_PROBE) 280 goto out; 281 282 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev); 283 284 skb = sock_alloc_send_skb(sk, length+hh_len+15, 285 flags&MSG_DONTWAIT, &err); 286 if (skb == NULL) 287 goto error; 288 skb_reserve(skb, hh_len); 289 290 skb->priority = sk->sk_priority; 291 skb->dst = dst_clone(&rt->u.dst); 292 293 skb_reset_network_header(skb); 294 iph = ip_hdr(skb); 295 skb_put(skb, length); 296 297 skb->ip_summed = CHECKSUM_NONE; 298 299 skb->transport_header = skb->network_header; 300 err = memcpy_fromiovecend((void *)iph, from, 0, length); 301 if (err) 302 goto error_fault; 303 304 /* We don't modify invalid header */ 305 if (length >= sizeof(*iph) && iph->ihl * 4U <= length) { 306 if (!iph->saddr) 307 iph->saddr = rt->rt_src; 308 iph->check = 0; 309 iph->tot_len = htons(length); 310 if (!iph->id) 311 ip_select_ident(iph, &rt->u.dst, NULL); 312 313 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl); 314 } 315 316 err = NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, skb, NULL, rt->u.dst.dev, 317 dst_output); 318 if (err > 0) 319 err = inet->recverr ? net_xmit_errno(err) : 0; 320 if (err) 321 goto error; 322out: 323 return 0; 324 325error_fault: 326 err = -EFAULT; 327 kfree_skb(skb); 328error: 329 IP_INC_STATS(IPSTATS_MIB_OUTDISCARDS); 330 return err; 331} 332 333static int raw_probe_proto_opt(struct flowi *fl, struct msghdr *msg) 334{ 335 struct iovec *iov; 336 u8 __user *type = NULL; 337 u8 __user *code = NULL; 338 int probed = 0; 339 unsigned int i; 340 341 if (!msg->msg_iov) 342 return 0; 343 344 for (i = 0; i < msg->msg_iovlen; i++) { 345 iov = &msg->msg_iov[i]; 346 if (!iov) 347 continue; 348 349 switch (fl->proto) { 350 case IPPROTO_ICMP: 351 /* check if one-byte field is readable or not. */ 352 if (iov->iov_base && iov->iov_len < 1) 353 break; 354 355 if (!type) { 356 type = iov->iov_base; 357 /* check if code field is readable or not. */ 358 if (iov->iov_len > 1) 359 code = type + 1; 360 } else if (!code) 361 code = iov->iov_base; 362 363 if (type && code) { 364 if (get_user(fl->fl_icmp_type, type) || 365 get_user(fl->fl_icmp_code, code)) 366 return -EFAULT; 367 probed = 1; 368 } 369 break; 370 default: 371 probed = 1; 372 break; 373 } 374 if (probed) 375 break; 376 } 377 return 0; 378} 379 380static int raw_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 381 size_t len) 382{ 383 struct inet_sock *inet = inet_sk(sk); 384 struct ipcm_cookie ipc; 385 struct rtable *rt = NULL; 386 int free = 0; 387 __be32 daddr; 388 __be32 saddr; 389 u8 tos; 390 int err; 391 392 err = -EMSGSIZE; 393 if (len > 0xFFFF) 394 goto out; 395 396 /* 397 * Check the flags. 398 */ 399 400 err = -EOPNOTSUPP; 401 if (msg->msg_flags & MSG_OOB) /* Mirror BSD error message */ 402 goto out; /* compatibility */ 403 404 /* 405 * Get and verify the address. 406 */ 407 408 if (msg->msg_namelen) { 409 struct sockaddr_in *usin = (struct sockaddr_in*)msg->msg_name; 410 err = -EINVAL; 411 if (msg->msg_namelen < sizeof(*usin)) 412 goto out; 413 if (usin->sin_family != AF_INET) { 414 static int complained; 415 if (!complained++) 416 printk(KERN_INFO "%s forgot to set AF_INET in " 417 "raw sendmsg. Fix it!\n", 418 current->comm); 419 err = -EAFNOSUPPORT; 420 if (usin->sin_family) 421 goto out; 422 } 423 daddr = usin->sin_addr.s_addr; 424 /* ANK: I did not forget to get protocol from port field. 425 * I just do not know, who uses this weirdness. 426 * IP_HDRINCL is much more convenient. 427 */ 428 } else { 429 err = -EDESTADDRREQ; 430 if (sk->sk_state != TCP_ESTABLISHED) 431 goto out; 432 daddr = inet->daddr; 433 } 434 435 ipc.addr = inet->saddr; 436 ipc.opt = NULL; 437 ipc.oif = sk->sk_bound_dev_if; 438 439 if (msg->msg_controllen) { 440 err = ip_cmsg_send(msg, &ipc); 441 if (err) 442 goto out; 443 if (ipc.opt) 444 free = 1; 445 } 446 447 saddr = ipc.addr; 448 ipc.addr = daddr; 449 450 if (!ipc.opt) 451 ipc.opt = inet->opt; 452 453 if (ipc.opt) { 454 err = -EINVAL; 455 /* Linux does not mangle headers on raw sockets, 456 * so that IP options + IP_HDRINCL is non-sense. 457 */ 458 if (inet->hdrincl) 459 goto done; 460 if (ipc.opt->srr) { 461 if (!daddr) 462 goto done; 463 daddr = ipc.opt->faddr; 464 } 465 } 466 tos = RT_CONN_FLAGS(sk); 467 if (msg->msg_flags & MSG_DONTROUTE) 468 tos |= RTO_ONLINK; 469 470 if (MULTICAST(daddr)) { 471 if (!ipc.oif) 472 ipc.oif = inet->mc_index; 473 if (!saddr) 474 saddr = inet->mc_addr; 475 } 476 477 { 478 struct flowi fl = { .oif = ipc.oif, 479 .nl_u = { .ip4_u = 480 { .daddr = daddr, 481 .saddr = saddr, 482 .tos = tos } }, 483 .proto = inet->hdrincl ? IPPROTO_RAW : 484 sk->sk_protocol, 485 }; 486 if (!inet->hdrincl) { 487 err = raw_probe_proto_opt(&fl, msg); 488 if (err) 489 goto done; 490 } 491 492 security_sk_classify_flow(sk, &fl); 493 err = ip_route_output_flow(&rt, &fl, sk, 1); 494 } 495 if (err) 496 goto done; 497 498 err = -EACCES; 499 if (rt->rt_flags & RTCF_BROADCAST && !sock_flag(sk, SOCK_BROADCAST)) 500 goto done; 501 502 if (msg->msg_flags & MSG_CONFIRM) 503 goto do_confirm; 504back_from_confirm: 505 506 if (inet->hdrincl) 507 err = raw_send_hdrinc(sk, msg->msg_iov, len, 508 rt, msg->msg_flags); 509 510 else { 511 if (!ipc.addr) 512 ipc.addr = rt->rt_dst; 513 lock_sock(sk); 514 err = ip_append_data(sk, ip_generic_getfrag, msg->msg_iov, len, 0, 515 &ipc, rt, msg->msg_flags); 516 if (err) 517 ip_flush_pending_frames(sk); 518 else if (!(msg->msg_flags & MSG_MORE)) 519 err = ip_push_pending_frames(sk); 520 release_sock(sk); 521 } 522done: 523 if (free) 524 kfree(ipc.opt); 525 ip_rt_put(rt); 526 527out: 528 if (err < 0) 529 return err; 530 return len; 531 532do_confirm: 533 dst_confirm(&rt->u.dst); 534 if (!(msg->msg_flags & MSG_PROBE) || len) 535 goto back_from_confirm; 536 err = 0; 537 goto done; 538} 539 540static void raw_close(struct sock *sk, long timeout) 541{ 542 /* 543 * Raw sockets may have direct kernel refereneces. Kill them. 544 */ 545 ip_ra_control(sk, 0, NULL); 546 547 sk_common_release(sk); 548} 549 550/* This gets rid of all the nasties in af_inet. -DaveM */ 551static int raw_bind(struct sock *sk, struct sockaddr *uaddr, int addr_len) 552{ 553 struct inet_sock *inet = inet_sk(sk); 554 struct sockaddr_in *addr = (struct sockaddr_in *) uaddr; 555 int ret = -EINVAL; 556 int chk_addr_ret; 557 558 if (sk->sk_state != TCP_CLOSE || addr_len < sizeof(struct sockaddr_in)) 559 goto out; 560 chk_addr_ret = inet_addr_type(addr->sin_addr.s_addr); 561 ret = -EADDRNOTAVAIL; 562 if (addr->sin_addr.s_addr && chk_addr_ret != RTN_LOCAL && 563 chk_addr_ret != RTN_MULTICAST && chk_addr_ret != RTN_BROADCAST) 564 goto out; 565 inet->rcv_saddr = inet->saddr = addr->sin_addr.s_addr; 566 if (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST) 567 inet->saddr = 0; /* Use device */ 568 sk_dst_reset(sk); 569 ret = 0; 570out: return ret; 571} 572 573/* 574 * This should be easy, if there is something there 575 * we return it, otherwise we block. 576 */ 577 578static int raw_recvmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *msg, 579 size_t len, int noblock, int flags, int *addr_len) 580{ 581 struct inet_sock *inet = inet_sk(sk); 582 size_t copied = 0; 583 int err = -EOPNOTSUPP; 584 struct sockaddr_in *sin = (struct sockaddr_in *)msg->msg_name; 585 struct sk_buff *skb; 586 587 if (flags & MSG_OOB) 588 goto out; 589 590 if (addr_len) 591 *addr_len = sizeof(*sin); 592 593 if (flags & MSG_ERRQUEUE) { 594 err = ip_recv_error(sk, msg, len); 595 goto out; 596 } 597 598 skb = skb_recv_datagram(sk, flags, noblock, &err); 599 if (!skb) 600 goto out; 601 602 copied = skb->len; 603 if (len < copied) { 604 msg->msg_flags |= MSG_TRUNC; 605 copied = len; 606 } 607 608 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 609 if (err) 610 goto done; 611 612 sock_recv_timestamp(msg, sk, skb); 613 614 /* Copy the address. */ 615 if (sin) { 616 sin->sin_family = AF_INET; 617 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 618 sin->sin_port = 0; 619 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 620 } 621 if (inet->cmsg_flags) 622 ip_cmsg_recv(msg, skb); 623 if (flags & MSG_TRUNC) 624 copied = skb->len; 625done: 626 skb_free_datagram(sk, skb); 627out: 628 if (err) 629 return err; 630 return copied; 631} 632 633static int raw_init(struct sock *sk) 634{ 635 struct raw_sock *rp = raw_sk(sk); 636 637 if (inet_sk(sk)->num == IPPROTO_ICMP) 638 memset(&rp->filter, 0, sizeof(rp->filter)); 639 return 0; 640} 641 642static int raw_seticmpfilter(struct sock *sk, char __user *optval, int optlen) 643{ 644 if (optlen > sizeof(struct icmp_filter)) 645 optlen = sizeof(struct icmp_filter); 646 if (copy_from_user(&raw_sk(sk)->filter, optval, optlen)) 647 return -EFAULT; 648 return 0; 649} 650 651static int raw_geticmpfilter(struct sock *sk, char __user *optval, int __user *optlen) 652{ 653 int len, ret = -EFAULT; 654 655 if (get_user(len, optlen)) 656 goto out; 657 ret = -EINVAL; 658 if (len < 0) 659 goto out; 660 if (len > sizeof(struct icmp_filter)) 661 len = sizeof(struct icmp_filter); 662 ret = -EFAULT; 663 if (put_user(len, optlen) || 664 copy_to_user(optval, &raw_sk(sk)->filter, len)) 665 goto out; 666 ret = 0; 667out: return ret; 668} 669 670static int do_raw_setsockopt(struct sock *sk, int level, int optname, 671 char __user *optval, int optlen) 672{ 673 if (optname == ICMP_FILTER) { 674 if (inet_sk(sk)->num != IPPROTO_ICMP) 675 return -EOPNOTSUPP; 676 else 677 return raw_seticmpfilter(sk, optval, optlen); 678 } 679 return -ENOPROTOOPT; 680} 681 682static int raw_setsockopt(struct sock *sk, int level, int optname, 683 char __user *optval, int optlen) 684{ 685 if (level != SOL_RAW) 686 return ip_setsockopt(sk, level, optname, optval, optlen); 687 return do_raw_setsockopt(sk, level, optname, optval, optlen); 688} 689 690#ifdef CONFIG_COMPAT 691static int compat_raw_setsockopt(struct sock *sk, int level, int optname, 692 char __user *optval, int optlen) 693{ 694 if (level != SOL_RAW) 695 return compat_ip_setsockopt(sk, level, optname, optval, optlen); 696 return do_raw_setsockopt(sk, level, optname, optval, optlen); 697} 698#endif 699 700static int do_raw_getsockopt(struct sock *sk, int level, int optname, 701 char __user *optval, int __user *optlen) 702{ 703 if (optname == ICMP_FILTER) { 704 if (inet_sk(sk)->num != IPPROTO_ICMP) 705 return -EOPNOTSUPP; 706 else 707 return raw_geticmpfilter(sk, optval, optlen); 708 } 709 return -ENOPROTOOPT; 710} 711 712static int raw_getsockopt(struct sock *sk, int level, int optname, 713 char __user *optval, int __user *optlen) 714{ 715 if (level != SOL_RAW) 716 return ip_getsockopt(sk, level, optname, optval, optlen); 717 return do_raw_getsockopt(sk, level, optname, optval, optlen); 718} 719 720#ifdef CONFIG_COMPAT 721static int compat_raw_getsockopt(struct sock *sk, int level, int optname, 722 char __user *optval, int __user *optlen) 723{ 724 if (level != SOL_RAW) 725 return compat_ip_getsockopt(sk, level, optname, optval, optlen); 726 return do_raw_getsockopt(sk, level, optname, optval, optlen); 727} 728#endif 729 730static int raw_ioctl(struct sock *sk, int cmd, unsigned long arg) 731{ 732 switch (cmd) { 733 case SIOCOUTQ: { 734 int amount = atomic_read(&sk->sk_wmem_alloc); 735 return put_user(amount, (int __user *)arg); 736 } 737 case SIOCINQ: { 738 struct sk_buff *skb; 739 int amount = 0; 740 741 spin_lock_bh(&sk->sk_receive_queue.lock); 742 skb = skb_peek(&sk->sk_receive_queue); 743 if (skb != NULL) 744 amount = skb->len; 745 spin_unlock_bh(&sk->sk_receive_queue.lock); 746 return put_user(amount, (int __user *)arg); 747 } 748 749 default: 750#ifdef CONFIG_IP_MROUTE 751 return ipmr_ioctl(sk, cmd, (void __user *)arg); 752#else 753 return -ENOIOCTLCMD; 754#endif 755 } 756} 757 758struct proto raw_prot = { 759 .name = "RAW", 760 .owner = THIS_MODULE, 761 .close = raw_close, 762 .connect = ip4_datagram_connect, 763 .disconnect = udp_disconnect, 764 .ioctl = raw_ioctl, 765 .init = raw_init, 766 .setsockopt = raw_setsockopt, 767 .getsockopt = raw_getsockopt, 768 .sendmsg = raw_sendmsg, 769 .recvmsg = raw_recvmsg, 770 .bind = raw_bind, 771 .backlog_rcv = raw_rcv_skb, 772 .hash = raw_v4_hash, 773 .unhash = raw_v4_unhash, 774 .obj_size = sizeof(struct raw_sock), 775#ifdef CONFIG_COMPAT 776 .compat_setsockopt = compat_raw_setsockopt, 777 .compat_getsockopt = compat_raw_getsockopt, 778#endif 779}; 780 781#ifdef CONFIG_PROC_FS 782struct raw_iter_state { 783 int bucket; 784}; 785 786#define raw_seq_private(seq) ((struct raw_iter_state *)(seq)->private) 787 788static struct sock *raw_get_first(struct seq_file *seq) 789{ 790 struct sock *sk; 791 struct raw_iter_state* state = raw_seq_private(seq); 792 793 for (state->bucket = 0; state->bucket < RAWV4_HTABLE_SIZE; ++state->bucket) { 794 struct hlist_node *node; 795 796 sk_for_each(sk, node, &raw_v4_htable[state->bucket]) 797 if (sk->sk_family == PF_INET) 798 goto found; 799 } 800 sk = NULL; 801found: 802 return sk; 803} 804 805static struct sock *raw_get_next(struct seq_file *seq, struct sock *sk) 806{ 807 struct raw_iter_state* state = raw_seq_private(seq); 808 809 do { 810 sk = sk_next(sk); 811try_again: 812 ; 813 } while (sk && sk->sk_family != PF_INET); 814 815 if (!sk && ++state->bucket < RAWV4_HTABLE_SIZE) { 816 sk = sk_head(&raw_v4_htable[state->bucket]); 817 goto try_again; 818 } 819 return sk; 820} 821 822static struct sock *raw_get_idx(struct seq_file *seq, loff_t pos) 823{ 824 struct sock *sk = raw_get_first(seq); 825 826 if (sk) 827 while (pos && (sk = raw_get_next(seq, sk)) != NULL) 828 --pos; 829 return pos ? NULL : sk; 830} 831 832static void *raw_seq_start(struct seq_file *seq, loff_t *pos) 833{ 834 read_lock(&raw_v4_lock); 835 return *pos ? raw_get_idx(seq, *pos - 1) : SEQ_START_TOKEN; 836} 837 838static void *raw_seq_next(struct seq_file *seq, void *v, loff_t *pos) 839{ 840 struct sock *sk; 841 842 if (v == SEQ_START_TOKEN) 843 sk = raw_get_first(seq); 844 else 845 sk = raw_get_next(seq, v); 846 ++*pos; 847 return sk; 848} 849 850static void raw_seq_stop(struct seq_file *seq, void *v) 851{ 852 read_unlock(&raw_v4_lock); 853} 854 855static __inline__ char *get_raw_sock(struct sock *sp, char *tmpbuf, int i) 856{ 857 struct inet_sock *inet = inet_sk(sp); 858 __be32 dest = inet->daddr, 859 src = inet->rcv_saddr; 860 __u16 destp = 0, 861 srcp = inet->num; 862 863 sprintf(tmpbuf, "%4d: %08X:%04X %08X:%04X" 864 " %02X %08X:%08X %02X:%08lX %08X %5d %8d %lu %d %p", 865 i, src, srcp, dest, destp, sp->sk_state, 866 atomic_read(&sp->sk_wmem_alloc), 867 atomic_read(&sp->sk_rmem_alloc), 868 0, 0L, 0, sock_i_uid(sp), 0, sock_i_ino(sp), 869 atomic_read(&sp->sk_refcnt), sp); 870 return tmpbuf; 871} 872 873static int raw_seq_show(struct seq_file *seq, void *v) 874{ 875 char tmpbuf[129]; 876 877 if (v == SEQ_START_TOKEN) 878 seq_printf(seq, "%-127s\n", 879 " sl local_address rem_address st tx_queue " 880 "rx_queue tr tm->when retrnsmt uid timeout " 881 "inode"); 882 else { 883 struct raw_iter_state *state = raw_seq_private(seq); 884 885 seq_printf(seq, "%-127s\n", 886 get_raw_sock(v, tmpbuf, state->bucket)); 887 } 888 return 0; 889} 890 891static const struct seq_operations raw_seq_ops = { 892 .start = raw_seq_start, 893 .next = raw_seq_next, 894 .stop = raw_seq_stop, 895 .show = raw_seq_show, 896}; 897 898static int raw_seq_open(struct inode *inode, struct file *file) 899{ 900 struct seq_file *seq; 901 int rc = -ENOMEM; 902 struct raw_iter_state *s = kmalloc(sizeof(*s), GFP_KERNEL); 903 904 if (!s) 905 goto out; 906 rc = seq_open(file, &raw_seq_ops); 907 if (rc) 908 goto out_kfree; 909 910 seq = file->private_data; 911 seq->private = s; 912 memset(s, 0, sizeof(*s)); 913out: 914 return rc; 915out_kfree: 916 kfree(s); 917 goto out; 918} 919 920static const struct file_operations raw_seq_fops = { 921 .owner = THIS_MODULE, 922 .open = raw_seq_open, 923 .read = seq_read, 924 .llseek = seq_lseek, 925 .release = seq_release_private, 926}; 927 928int __init raw_proc_init(void) 929{ 930 if (!proc_net_fops_create("raw", S_IRUGO, &raw_seq_fops)) 931 return -ENOMEM; 932 return 0; 933} 934 935void __init raw_proc_exit(void) 936{ 937 proc_net_remove("raw"); 938} 939#endif /* CONFIG_PROC_FS */ 940